CN108090910A - It is a kind of that tomato plant image segmentation algorithm outside the night room of PCNN models is simplified based on comentropy gradient - Google Patents
It is a kind of that tomato plant image segmentation algorithm outside the night room of PCNN models is simplified based on comentropy gradient Download PDFInfo
- Publication number
- CN108090910A CN108090910A CN201810006510.8A CN201810006510A CN108090910A CN 108090910 A CN108090910 A CN 108090910A CN 201810006510 A CN201810006510 A CN 201810006510A CN 108090910 A CN108090910 A CN 108090910A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- formula
- max
- image segmentation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
Tomato plant image segmentation algorithm outside the night room of PCNN models is simplified based on comentropy gradient the invention discloses a kind of, aberration is normalized calculatingC n Afterwards, image segmentation iteration is carried out to normalization aberration gray-scale map using simplified PCNN models；Calculate image segmentation resultY(n) comentropys(n)；Whens(n) become smaller or iterationsnMaximum iteration is reachedMWhen, stop iteration；Seek maximum informational entropys(max) and it is correspondingmax；It calculatess(max) withs(max1) difference, i.e. comentropy gradientΔs；JudgeΔsWhether threshold value is more thanT s , if so, withmaxWhat+1 iteration obtainedY(max+ 1) it is image segmentation resultF, otherwise, withY(max) it is image segmentation resultF；It is rightFCarry out denoising.The present invention is using comentropy gradient as optimized image segmentation result interpretational criteria, use simplified PCNN models, uneven illumination is even under the conditions of overcoming night active illumination, there are the influences that shade, specular etc. split tomato plant image outside night room, realize tomato plant image outside night room and split.
Description
Technical field
The present invention relates to a kind of image processing methods, simplify PCNN models based on comentropy gradient more particularly, to a kind of
Tomato plant image segmentation algorithm outside night room.
Background technology
Tomato production robot is a solution for realizing tomato production automation, and the tomato based on machine vision is planted
It is that tomato production robot obtains tomato plant composition organ size and position that strain composition organ (including stem, leaf, tomato etc.), which identifies,
Put etc. information (for tomato production robot carry out automatic fruit thinning, dredge leaf, target sprayed, is picked, avoidance, navigation etc.) one kind
Means.It is to realize the premise of tomato plant composition organ identification to realize the segmentation of tomato plant image.To extend tomato production machine
The working time of people improves the utilization rate and work efficiency of tomato production robot, realizes tomato production robot in night ring
Automated job under border is highly desirable, and realizes that the segmentation of tomato plant image is to realize tomato production robot outside night room
Obtain the premise of the information such as tomato plant composition organ size and location automatically in night work.
There are following features for the outdoor tomato plant image gathered under the conditions of night active illumination：Different images gather distance
Under, illumination condition is there are larger difference, and there are larger differences for the corresponding brightness for gathering image；In same image, brightness disproportionation
Even, distribution of light sources center corresponding region is substantially partially bright compared with other regions；There are speculars towards light source one side for stalk, leaf, fruit.
These features cause bigger difficulty to the segmentation of night tomato plant image.
It is gathered under the conditions of fruits and vegetables production machine human visual system's research field, existing more natural lighting in the daytime green
Color segmentation of crop images algorithm.Since in the daytime and night, illumination condition is there are larger difference, therefore the green agriculture gathered in the daytime
Crop map picture shows the characteristics of different from the tomato plant image of night acquisition.In the daytime under the conditions of natural lighting, due to illumination
Variation is violent, and the image gathered under different illumination conditions, luminance difference is larger, and image is presented as that image segmentation threshold is poor when splitting
It is different larger；Uneven illumination is even, and the green crop area brightness for causing different position on image is different；Background is complicated.In the daytime certainly
The fruits and vegetables plant image difference gathered under right illumination condition and under the conditions of night active illumination is larger, causes image segmentation algorithm same
There are larger differences for sample.Therefore, for the image segmentation algorithm of the green crop map picture gathered under the conditions of natural lighting in the daytime
It cannot be directly used to the image segmentation of the tomato plant image gathered under the conditions of night active illumination.
Currently, existing night fruits and vegetables production machine human visual system's image segmentation algorithm is mainly night fruit image point
Algorithm is cut, the fruit tree image segmentation algorithm for separately there are the branches and leaves under the conditions of a small number of night active illuminations heterochromatic.Tomato plant and fruit
Vegetable trunk, branch and fruit optical characteristics are different, and are had differences in color, size etc., therefore, currently existing night
Fruit and the heterochromatic fruit tree image segmentation algorithm of branches and leaves cannot be directly used to the tomato plant image segmentation of night branches and leaves advancing coloud nearside.
The currently existing branches and leaves advancing coloud nearside fruits and vegetables plant image segmentation algorithm for being suitable for gathering under the conditions of night active illumination
Seldom, and the illumination variation under is not gathered for different images, uneven illumination is even, and there are the nights fruits such as shade, specular
The characteristics of vegetable plant image, carries out the design of image segmentation algorithm, and image segmentation can not still meet the needs of production practices.
The content of the invention
Tomato plant outside the night room of PCNN models is simplified based on comentropy gradient it is an object of the invention to provide a kind of
Image segmentation algorithm, using criterion of the comentropy gradient as optimized image segmentation result, using a kind of simplified PCNN moulds
Type realizes the image segmentation of tomato plant outside night room, lays the foundation for the identification and positioning of tomato plant organ outside night room.
The technical solution adopted by the present invention is：
Tomato plant outside night room is illuminated using lighting system；Image received device receives active illumination condition
Under night room outside tomato plant optical imagery after, be converted into electronic image output；The electricity of image received device output
Subgraph is input to image analysis processing system；Image analysis processing system uses night to Tomato Image outside the night room of input
Outdoor tomato plant image segmentation algorithm realizes tomato plant image segmentation outside night room；It is actively shone with color camera acquisition night
The coloured image C of outdoor tomato plant under the conditions of bright reads R, G, B color component of each pixel in coloured image C；Including following
Step：
1. normalize value of chromatism CnIt calculates：Each pixel on coloured image C is calculated respectively by formula (1) green red
Aberration Cc, green red difference C is calculatedcAfterwards, normalization value of chromatism C is calculated by formula (2)n, by CnAs pixel value, ash is obtained
Spend image I；
In formula (2), min (ccThe c of all pixels point on)-coloured image CcMinimum value；max(cc)-coloured image C
The c of upper all pixels pointcMaximum.
2. it is 1 to set iterations n；
3. carrying out image segmentation to I using PCNN models are simplified, image segmentation result Y (n) is obtained；Simplify PCNN models,
It is i by image abscissa, ordinate is that the pixel of j is considered as a neuron Nij；By each pixel in gray level image I pixel value
IijInput as neuron；The output of neuron is Yij(n) (in the image segmentation result Y (n) that i.e. nth iteration obtains,
Abscissa is i in image coordinate system, and ordinate is the pixel value of the pixel of j) it is calculated by formula (3)；All neurons
Output is the image segmentation result Y (n) that nth iteration obtains；
In formula (3), n-nth iteration；U* ij(n)-internal activity item, as shown in formula (4), Eij(n)-dynamic threshold,
As shown in formula (5)；Yij(0) it is 0；
In formula (4), Fij(n)-neuron input item, as shown in formula (6)；L* ij(n)-connection input item, such as formula (7) institute
Show；β-internal activity item coefficient of connection, is set to 0.1；
Eij(n)=T+-n×N n≤M (5)
In formula (5), T+- dynamic threshold initial value, is set to 0.45；N-threshold value attenuation steps, is set to 0.025；M-maximum
Iterations is 8 as shown in formula (8)；Eij(0) it is T+；
Fij(n)=Iij (6)
In formula (6), Iij- external input encourages, i.e. image pixel value；
In formula (7), Lij(n) as shown in formula (9)；
In formula (8), K-dynamic threshold constant interval width is set to 0.2；
Lij(n)=∑ WijYij(n-1) (9)
In formula (9), Wij- be of coupled connections domain coefficient of connection, as shown in formula (10),
Wij=(1- | Fij(n)-Fi+k,j+l(n)|)e1-d (10)
In formula (10)：E-Euler's numbers, is approximately equal to 2.718281828；D-neuron NklWith neuron NijBetween it is European
Distance is calculated as shown in formula (11),
In formula (11)：xi+k, yj+l- neuron NklCoordinate；xi, yj- neuron NijCoordinate；
4. calculating the comentropy s (n) of image segmentation result Y (n), judge whether n is 1, if so, by maximum informational entropy s
(max) value is revised as s (1), and the value of max is revised as 1, is transferred to step 6.；Otherwise, it is transferred to step 5.；
5. judge whether s (n) is more than maximum informational entropy s (max), if so, the value of maximum informational entropy s (max) is changed
For s (n), and the value of max is revised as n, is transferred to step 6.；Otherwise, terminate iterative process, go to step 7.；
6. judging whether n is equal to default maximum iteration M, if so, terminating iterative process, step is transferred to 7.；It is no
Then, 3. iterations n jumps to step from increasing 1；
7. calculating the difference for the comentropy s (max-1) that maximum informational entropy s (max) is obtained with the max-1 times iteration, that is, believe
Cease entropy gradient delta s；
8. judge whether Δ s is more than threshold value Ts(being set to 0.17), if so, the Y (max+ obtained with the max+1 times iteration
1) it is image segmentation result F；Otherwise, the Y (max) obtained with the corresponding the max times iteration of maximum informational entropy is image segmentation knot
Fruit F；
9. image segmentation result denoising：It removes area in image segmentation result F and is less than predetermined threshold value TaRegion, gone
Image segmentation result F after making an uproarn。
The invention has the advantages that：
The present invention using comentropy gradient as optimized image segmentation result interpretational criteria, using simplified PCNN models,
Different images gather the illumination variation under under the conditions of overcoming night active illumination, and uneven illumination is even, and there are shade, blooms
The influence that tomato plant image outside night room is split in area etc. realizes under the conditions of active illumination tomato plant image outside night room
Segmentation is laid a good foundation for the identification and positioning of tomato plant organ outside night room.
Description of the drawings
Fig. 1 is tomato plant image segmentation system schematic diagram outside night room.
Fig. 2 is to simplify tomato plant image segmentation algorithm flow chart outside the night room of PCNN models based on comentropy gradient.
Fig. 3 is simplified PCNN illustratons of model.
In Fig. 1：1st, tomato plant, 2, color camera, 3, lighting system, 4,1394 image pick-up cards, 5, computer, 6, night
Between outdoor Tomato Image Segmentation software.
Specific embodiment
The invention will be further described with reference to the accompanying drawings and examples.
Fig. 1 illustrates a specific embodiment of tomato plant image segmentation system outside night room.Lighting system 3 uses 2
The diagonal irradiation system that the white fluorescent lamp of small cup 3w is formed, diagonal distance 400mm.Image received device uses color camera 2,
Imaging sensor in color camera 2 is colour Sony ICX204 CCD, and maximum resolution is 1024 × 768, and the focal length of lens is
6mm.4 model MOGE 1394 of image pick-up card, charged source adapter.Computer 5 is Lenovo R400 laptops, interior
3G is deposited, CPU is 7 operating system of Intel CoreDuo T6570, WIN.Using 1394 connecting lines by 2 and 1394 figure of color camera
Picture capture card 4 is connected, and 1394 image pick-up cards 4 close 1 card reader interfaces by 7 and are mounted on computer 5.
What tomato plant image was split outside night room is implemented as follows：
Tomato plant 1 outside night room is illuminated using lighting system 3；Color camera 2 receives the light of tomato plant 1
After learning image, electronic image output is converted into；The electronic image that color camera 2 exports is input to 1394 image pick-up cards 4；
Analog picture signal is converted to the outdoor tomato being input to after data image signal in computer 5 and planted by 1394 image pick-up cards 4
Strain image segmentation software 6；Outdoor tomato plant image segmentation software 6 realizes night using outdoor tomato plant image partition method
Tomato plant image is split under the conditions of active illumination.
Coloured image C (640 × 480 pictures of outdoor tomato plant under the conditions of night active illumination are gathered with color camera
Element), read R, G, B color component of each pixel in coloured image C；As shown in Fig. 2, comprise the following steps：
1. normalize value of chromatism CnIt calculates：Each pixel on coloured image C is calculated respectively by formula (1) green red
Aberration Cc, green red difference C is calculatedcAfterwards, normalization value of chromatism C is calculated by formula (2)n, by CnAs pixel value, ash is obtained
Spend image I；
In formula (2), min (ccThe c of all pixels point on)-coloured image CcMinimum value；max(cc)-coloured image C
The c of upper all pixels pointcMaximum.
2. it is 1 to set iterations n；
3. carrying out image segmentation to I using PCNN models are simplified, image segmentation result Y (n) is obtained；Simplify PCNN models,
As shown in figure 3, including three parts：Coupling input area；Internal activity domain；Impulse generator.By image abscissa be i, ordinate
It is considered as a neuron N for the pixel of jij；By each pixel in gray level image I pixel value IijInput as neuron；Nerve
The output of member is Yij(n) (in the image segmentation result Y (n) that i.e. nth iteration obtains, abscissa is i in image coordinate system, is indulged
Coordinate is the pixel value of the pixel of j) it is calculated by formula (3)；The output of all neurons is the figure that nth iteration obtains
As segmentation result Y (n)；
In formula (3), n-nth iteration；U* ij(n)-internal activity item, the output in internal activity domain, as shown in formula (4),
Eij(n)-dynamic threshold is obtained using linear threshold attenuator, as shown in formula (5)；Yij(0) it is 0；
In formula (4), Fij(n)-neuron input item, as shown in formula (6)；L* ij(n)-connection input item, such as formula (7) institute
Show；β-internal activity item coefficient of connection, is set to 0.1；
Eij(n)=T+-n×N n≤M (5)
In formula (5), T+- dynamic threshold initial value, is set to 0.45；N-threshold value attenuation steps, is set to 0.025；M-maximum
Iterations is 8 as shown in formula (8)；Eij(0) it is T+；
Fij(n)=Iij (6)
In formula (6), Iij- external input encourages, i.e. image pixel value；
In formula (7), Lij(n) as shown in formula (9)；
In formula (8), K-dynamic threshold constant interval width is set to 0.2；
Lij(n)=∑ WijYij(n-1) (9)
In formula (9), Wij- be of coupled connections domain coefficient of connection, as shown in formula (10),
Wij=(1- | Fij(n)-Fi+k,j+l(n)|)e1-d (10)
In formula (10)：E-Euler's numbers, is approximately equal to 2.718281828；D-neuron NklWith neuron NijBetween it is European
Distance is calculated as shown in formula (11),
In formula (11)：xi+k, yj+l- neuron NklCoordinate；xi, yj- neuron NijCoordinate；
4. the comentropy s (n) of image segmentation result Y (n) is calculated, shown in the calculating such as formula (12) of comentropy；Whether judge n
For 1, if so, the value of maximum informational entropy s (max) is revised as s (1), and the value of max is revised as 1, is transferred to step 6.；It is no
Then, it is transferred to step 5.；
In formula (12), pi(n) shown in calculating such as formula (13),
In formula (13), Ti(n) pixel value is the pixel number of i in-Y (n)；Sum of all pixels in T (n)-Y (n)；
5. judge whether s (n) is more than maximum informational entropy s (max), if so, the value of maximum informational entropy s (max) is changed
For s (n), and the value of max is revised as n, is transferred to step 6.；Otherwise, terminate iterative process, go to step 7.；
6. judging whether n is equal to default maximum iteration M, if so, terminating iterative process, step is transferred to 7.；It is no
Then, 3. iterations n jumps to step from increasing 1；
7. calculating the difference for the comentropy s (max-1) that maximum informational entropy s (max) is obtained with the max-1 times iteration, that is, believe
Cease entropy gradient delta s；
8. judge whether Δ s is more than threshold value Ts(being set to 0.17), if so, the Y (max+ obtained with the max+1 times iteration
1) it is image segmentation result F；Otherwise, the Y (max) obtained with the corresponding the max times iteration of maximum informational entropy is image segmentation knot
Fruit F；
9. image segmentation result denoising：It removes area in image segmentation result F and is less than predetermined threshold value Ta(being set to 300 pixels)
Region, obtain the image segmentation result F after denoisingn。
Claims (1)
1. a kind of simplify tomato plant image segmentation algorithm outside the night room of PCNN models based on comentropy gradient, color camera is used
The coloured image C of outdoor tomato plant under the conditions of acquisition night active illumination reads R, G, B face of each pixel in coloured image C
Colouring component；It is characterised in that it includes following steps：
1. normalize value of chromatism CnIt calculates：It is poor that green red is calculated by formula (1) respectively to each pixel on coloured image C
Cc, green red difference C is calculatedcAfterwards, normalization value of chromatism C is calculated by formula (2)n, by CnAs pixel value, gray-scale map is obtained
As I；
<mrow>
<msub>
<mi>c</mi>
<mi>c</mi>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<mn>2</mn>
<mi>G</mi>
<mo>-</mo>
<mi>R</mi>
<mo>-</mo>
<mi>B</mi>
</mrow>
<mrow>
<mi>R</mi>
<mo>+</mo>
<mi>G</mi>
<mo>+</mo>
<mi>B</mi>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>c</mi>
<mi>n</mi>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>c</mi>
<mi>c</mi>
</msub>
<mo>-</mo>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>c</mi>
<mi>c</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>max</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>c</mi>
<mi>c</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>min</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>c</mi>
<mi>c</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula (2), min (ccThe c of all pixels point on)-coloured image CcMinimum value；max(ccInstitute on)-coloured image C
There is the c of pixelcMaximum.
2. it is 1 to set iterations n；
3. carrying out image segmentation to I using PCNN models are simplified, image segmentation result Y (n) is obtained；Simplify the feature of PCNN models
It is：It is i by image abscissa, ordinate is that the pixel of j is considered as a neuron Nij；By each pixel in gray level image I picture
Plain value IijInput as neuron；The output of neuron is Yij(n) (the image segmentation result Y (n) that i.e. nth iteration obtains
In, abscissa is i in image coordinate system, and ordinate is the pixel value of the pixel of j) it is calculated by formula (3)；All neurons
Output be the obtained image segmentation result Y (n) of nth iteration；
<mrow>
<msub>
<mi>Y</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mrow>
<msubsup>
<mi>U</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mo>*</mo>
</msubsup>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>></mo>
<msub>
<mi>E</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>-</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mrow>
<msubsup>
<mi>U</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mo>*</mo>
</msubsup>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>&le;</mo>
<msub>
<mi>E</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>-</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula (3), n-nth iteration；U* ij(n)-internal activity item, as shown in formula (4), Eij(n)-dynamic threshold, such as formula
(5) shown in；Yij(0) it is 0；
<mrow>
<msubsup>
<mi>U</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mo>*</mo>
</msubsup>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msub>
<mi>F</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<msubsup>
<mi>&beta;L</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mo>*</mo>
</msubsup>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula (4), Fij(n)-neuron input item, as shown in formula (6)；L* ij(n)-connection input item, as shown in formula (7)；
β-internal activity item coefficient of connection, is set to 0.1；
Eij(n)=T+-n×Nn≤M (5)
In formula (5), T+- dynamic threshold initial value, is set to 0.45；N-threshold value attenuation steps, is set to 0.025；M-greatest iteration
Number is 8 as shown in formula (8)；Eij(0) it is T+；
Fij(n)=Iij (6)
In formula (6), Iij- external input encourages, i.e. image pixel value；
<mrow>
<msubsup>
<mi>L</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mo>*</mo>
</msubsup>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mrow>
<msub>
<mi>L</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>></mo>
<mn>0</mn>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mrow>
<msub>
<mi>L</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>&le;</mo>
<mn>0</mn>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>7</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula (7), Lij(n) as shown in formula (9)；
<mrow>
<mi>M</mi>
<mo>=</mo>
<mfrac>
<mi>K</mi>
<mi>N</mi>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>8</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula (8), K-dynamic threshold constant interval width is set to 0.2；
Lij(n)=∑ WijYij(n-1) (9)
In formula (9), Wij- be of coupled connections domain coefficient of connection, as shown in formula (10),
Wij=(1- | Fij(n)-Fi+k,j+l(n)|)e1-d (10)
In formula (10)：E-Euler's numbers, is approximately equal to 2.718281828；D-neuron NklWith neuron NijBetween Euclidean distance,
It is calculated as shown in formula (11),
<mrow>
<mi>d</mi>
<mo>=</mo>
<msqrt>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mrow>
<mi>i</mi>
<mo>+</mo>
<mi>k</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mrow>
<mi>j</mi>
<mo>+</mo>
<mi>l</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>11</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula (11)：xi+k, yj+l- neuron NklCoordinate；xi, yj- neuron NijCoordinate；
4. calculating the comentropy s (n) of image segmentation result Y (n), judge whether n is 1, if so, by maximum informational entropy s (max)
Value be revised as s (1), and the value of max is revised as 1, is transferred to step 6.；Otherwise, it is transferred to step 5.；
5. judge whether s (n) is more than maximum informational entropy s (max), if so, the value of maximum informational entropy s (max) is revised as s
(n), n and by the value of max is revised as, is transferred to step 6.；Otherwise, terminate iterative process, go to step 7.；
6. judging whether n is equal to default maximum iteration M, if so, terminating iterative process, step is transferred to 7.；Otherwise, repeatedly
For frequency n from increasing 1, step is jumped to 3.；
7. calculate the difference for the comentropy s (max-1) that maximum informational entropy s (max) is obtained with the max-1 times iteration, i.e. comentropy
Gradient delta s；
8. judge whether Δ s is more than threshold value Ts, if so, the Y (max+1) obtained with the max+1 times iteration is image segmentation result
F；Otherwise, the Y (max) obtained with the corresponding the max times iteration of maximum informational entropy is image segmentation result F；
9. image segmentation result denoising：It removes area in image segmentation result F and is less than predetermined threshold value TaRegion, after obtaining denoising
Image segmentation result Fn。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810006510.8A CN108090910B (en) | 2018-01-04 | 2018-01-04 | Night outdoor tomato plant image segmentation algorithm based on information entropy gradient simplified PCNN model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810006510.8A CN108090910B (en) | 2018-01-04 | 2018-01-04 | Night outdoor tomato plant image segmentation algorithm based on information entropy gradient simplified PCNN model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108090910A true CN108090910A (en) | 2018-05-29 |
| CN108090910B CN108090910B (en) | 2021-04-20 |
Family
ID=62179968
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810006510.8A Expired - Fee Related CN108090910B (en) | 2018-01-04 | 2018-01-04 | Night outdoor tomato plant image segmentation algorithm based on information entropy gradient simplified PCNN model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108090910B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108901540A (en) * | 2018-06-28 | 2018-11-30 | 重庆邮电大学 | Fruit tree light filling and fruit thinning method based on artificial bee colony fuzzy clustering algorithm |
| CN109255795A (en) * | 2018-09-11 | 2019-01-22 | 中国计量大学 | A kind of tomato plant edge sort algorithm |
| CN110889876A (en) * | 2019-12-10 | 2020-03-17 | 兰州交通大学 | Color image quantization method based on CA-SPCNN algorithm |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103177445A (en) * | 2013-03-13 | 2013-06-26 | 浙江大学 | Outdoor tomato identification method based on subsection threshold image segmentation and light spot identification |
| US20150134583A1 (en) * | 2013-11-14 | 2015-05-14 | Denso Corporation | Learning apparatus, learning program, and learning method |
| CN105005975A (en) * | 2015-07-08 | 2015-10-28 | 南京信息工程大学 | Image de-noising method based on anisotropic diffusion of image entropy and PCNN |
| CN106023224A (en) * | 2016-05-30 | 2016-10-12 | 天水师范学院 | PCNN automatic segmentation method for microscopic image of traditional Chinese medicine |
| CN106067026A (en) * | 2016-05-30 | 2016-11-02 | 天水师范学院 | A kind of Feature extraction and recognition search method of microimage of Chinese medical herb |
| CN107016676A (en) * | 2017-03-13 | 2017-08-04 | 三峡大学 | A kind of retinal vascular images dividing method and system based on PCNN |
-
2018
- 2018-01-04 CN CN201810006510.8A patent/CN108090910B/en not_active Expired - Fee Related
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103177445A (en) * | 2013-03-13 | 2013-06-26 | 浙江大学 | Outdoor tomato identification method based on subsection threshold image segmentation and light spot identification |
| CN103177445B (en) * | 2013-03-13 | 2015-10-28 | 浙江大学 | Based on the outdoor tomato recognition methods of fragmentation threshold Iamge Segmentation and spot identification |
| US20150134583A1 (en) * | 2013-11-14 | 2015-05-14 | Denso Corporation | Learning apparatus, learning program, and learning method |
| CN105005975A (en) * | 2015-07-08 | 2015-10-28 | 南京信息工程大学 | Image de-noising method based on anisotropic diffusion of image entropy and PCNN |
| CN106023224A (en) * | 2016-05-30 | 2016-10-12 | 天水师范学院 | PCNN automatic segmentation method for microscopic image of traditional Chinese medicine |
| CN106067026A (en) * | 2016-05-30 | 2016-11-02 | 天水师范学院 | A kind of Feature extraction and recognition search method of microimage of Chinese medical herb |
| CN107016676A (en) * | 2017-03-13 | 2017-08-04 | 三峡大学 | A kind of retinal vascular images dividing method and system based on PCNN |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108901540A (en) * | 2018-06-28 | 2018-11-30 | 重庆邮电大学 | Fruit tree light filling and fruit thinning method based on artificial bee colony fuzzy clustering algorithm |
| CN109255795A (en) * | 2018-09-11 | 2019-01-22 | 中国计量大学 | A kind of tomato plant edge sort algorithm |
| CN109255795B (en) * | 2018-09-11 | 2021-04-06 | 中国计量大学 | Tomato plant edge sorting method |
| CN110889876A (en) * | 2019-12-10 | 2020-03-17 | 兰州交通大学 | Color image quantization method based on CA-SPCNN algorithm |
| CN110889876B (en) * | 2019-12-10 | 2022-05-03 | 兰州交通大学 | Color image quantization method based on CA-SPCNN algorithm |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108090910B (en) | 2021-04-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gan et al. | Immature green citrus fruit detection using color and thermal images | |
| RU2735151C2 (en) | Weeds identification in natural environment | |
| Liu et al. | A method of segmenting apples at night based on color and position information | |
| Nuske et al. | Automated visual yield estimation in vineyards | |
| CN110148120A (en) | A kind of disease intelligent identification Method and system based on CNN and transfer learning | |
| Song et al. | Detection of maize tassels for UAV remote sensing image with an improved YOLOX model | |
| CN107346434A (en) | A kind of plant pest detection method based on multiple features and SVMs | |
| US11455794B2 (en) | System and method for orchard recognition on geographic area | |
| Liu et al. | Detection of citrus fruit and tree trunks in natural environments using a multi-elliptical boundary model | |
| CN109325495A (en) | A kind of crop image segmentation system and method based on deep neural network modeling | |
| CN103177445B (en) | Based on the outdoor tomato recognition methods of fragmentation threshold Iamge Segmentation and spot identification | |
| EP3279831A1 (en) | Recognition of weed in a natural environment using a digital image | |
| CN110070571B (en) | Phyllostachys pubescens morphological parameter detection method based on depth camera | |
| KR102002632B1 (en) | Fruit monitoring system and method at the same | |
| CN108090910A (en) | It is a kind of that tomato plant image segmentation algorithm outside the night room of PCNN models is simplified based on comentropy gradient | |
| CN106815826A (en) | Night vision image Color Fusion based on scene Recognition | |
| Çakır et al. | Detection of oranges in outdoor conditions | |
| CN113252584B (en) | Crop growth detection method and system based on 5G transmission | |
| CN114612794A (en) | Remote sensing identification method for land covering and planting structure in finely-divided agricultural area | |
| Liu et al. | Development of a machine vision algorithm for recognition of peach fruit in a natural scene | |
| Zhang et al. | Yield estimation of citrus fruit using rapid image processing in natural background | |
| Liu et al. | Recognition and localization of actinidia arguta based on image recognition | |
| CN115861686A (en) | Litchi key growth period identification and detection method and system based on edge deep learning | |
| Zhang et al. | Recognition of greenhouse cucumber fruit using computer vision | |
| CN103900498B (en) | A kind of cotton field automatic detection method of the growth of cereal crop seedlings and detection device thereof |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210420 Termination date: 20220104 |